Development of a mathematical model to analyze color and density as discriminant features for pulmonary squamous epithelial cells

摘要

A new mathematical technique, originally developed to analyze multispectral infrared scans of the earth collected from orbiting satellites has been applied to differentiating cell types encountered during lung squamous cell carcinogenesis. Cell digital images were collected at wavelengths of 430 nm, 530 nm, and 590 nm using a Zeiss scanning microscope photometer system.A mathematical model to measure relative color and density changes was developed. Color and density features were generated for each cell based on the transformed data. Separation statistics, correlation studies, and classification accuracies provide encouraging preliminary results for the use of the mathematical model.The results in this paper are restricted to non-orangeophilic cells. A group of 201 cells from seven subject populations ranging in type from metaplastic to malignant were analyzed using automated computer generated features.